Re: libc_fp* entries for mips debug set?
<[email protected]> Fri, 21 Oct 2016 18:24:31 +0000
| Newsgroups | gmane.os.netbsd.ports.evbmips,gmane.os.netbsd.current,gmane.os.netbsd.ports.mips.devel |
|---|---|
| Message-ID | <[email protected]> |
"tf" means "long double", which I didn't think MIPS supports. So this look= s like a case of some bit of code thinking that it does when it actually is= n't there. paul > On Oct 21, 2016, at 2:16 PM, Michael <[email protected]> wrote: >=20 > Hello, >=20 > On Fri, 21 Oct 2016 12:27:54 +0200 > Martin Husemann <[email protected]> wrote: >=20 >> On Fri, Oct 21, 2016 at 05:18:44AM -0500, John D. Baker wrote: >>> This is evbmips-mips64el for my Lemote YeeLoong. >>=20 >> So that should be arch64, but not softfloat, and I would have expected >> the entries to be ignored. >=20 > All mips64 is softfloat for now, libc_fp just implements some of it > using FPU instructions. Then again, I wouldn't be surprised if I goofed > the set list attributes for those files... >=20 > That said, hardfloat builds currently fail with > /home/build/dest_evbmips/lib/libc.so: undefined reference to `__gttf2' > /home/build/dest_evbmips/lib/libc.so: undefined reference to `__divtf3' > /home/build/dest_evbmips/lib/libc.so: undefined reference to `__multf3' > /home/build/dest_evbmips/lib/libc.so: undefined reference to `__floatunsi= tf' >=20 > probably fallout from moving the softfloat stuff from libc to libgcc > recently? >=20 > have fun > Michael